To provide quick responses, enterprises usually build large data centers (IDC, Internet Data Center) in multiple regions and each data center stores copies of data, in order to reduce the forwarding path of user requests. The synchronization and update of data among the data centers become increasingly prominent key issue. Currently in the prior art, data transmission between the data centers is generally implemented using the peer-to-peer (P2P) technology. A tracking server is used to record the data status and content currently being downloaded by each server (peer) in the data centers, to query, for each server that needs to download data, a list of other servers storing the data requested by that server, and to randomly select a server from the list of other servers as the source server for the data transmission, so as to complete the data transmission.
However, the prior art selects the source server at random, fails to select the most appropriate source server from a holistic perspective, cannot ensure that the data transmission can be completed before a given time, and cannot effectively utilize the link bandwidth.